Linux Fix Queued For Recent AMD Laptops Failing To Resume From Attached USB Devices

[phoronix]

Phoronix: Linux Fix Queued For Recent AMD Laptops Failing To Resume When USB Devices Attached

A fix has made its way into the Linux PCI subsystem's power management branch to address various AMD Ryzen Rembrandt and Phoenix generation laptops failing to resume from suspend when external USB devices are attached...

https://www.phoronix.com/news/Linux-Fix-AMD-Laptop-Resume-USB

[Zajec]

I don't think it's about "Resume When USB Devices Attached" but rather about resume using attached USB devices like USB keyboard. Please verify I read it right though.

[Michael]

I don't think it's about "Resume When USB Devices Attached" but rather about resume using attached USB devices like USB keyboard. Please verify I read it right though.

Yeah, cleared up the title, thanks.

[yump]

From the bug report and commit message, it sounds like the symptom wasn't that the laptop couldn't resume, but rather that a USB 4-attached keyboard couldn't cause it to resume. But it could still be woken by pressing the power button (and maybe other methods that you'd use with no external keyboard, like the internal keyboard or lid switch.

The fix was to not put that USB controller to sleep during S0ix suspend, which could be trading battery life for correctness in an uncommon situation, and would contribute to the ongoing embarrassment of S0ix by old school S3. It seems to me that this should be a runtime-comfortable workaround, in case users care more about suspended battery drain than the ability to resume by external keyboard connected via USB 4.

[superm1]

From the bug report and commit message, it sounds like the symptom wasn't that the laptop couldn't resume, but rather that a USB 4-attached keyboard couldn't cause it to resume. But it could still be woken by pressing the power button (and maybe other methods that you'd use with no external keyboard, like the internal keyboard or lid switch.

The fix was to not put that USB controller to sleep during S0ix suspend, which could be trading battery life for correctness in an uncommon situation, and would contribute to the ongoing embarrassment of S0ix by old school S3. It seems to me that this should be a runtime-comfortable workaround, in case users care more about suspended battery drain than the ability to resume by external keyboard connected via USB 4.

The USB4 router is put into D3 still, this changes the behavior of the root port that it's connected to however in preventing it from going into D3 when you have a wakeup source programmed downstream. If you don't want to be able to wake up from a downstream USB device you can change the "power/wakeup" attribute in sysfs for that device. This quirk will obey it.

[yump]

The USB4 router is put into D3 still, this changes the behavior of the root port that it's connected to however in preventing it from going into D3 when you have a wakeup source programmed downstream. If you don't want to be able to wake up from a downstream USB device you can change the "power/wakeup" attribute in sysfs for that device. This quirk will obey it.

So the battery drain will only occur when a USB keyboard is actually connected, correct? External keyboard without external power seems like a rare situation, so it probably won't cause much trouble.

[superm1]

So the battery drain will only occur when a USB keyboard is actually connected, correct? External keyboard without external power seems like a rare situation, so it probably won't cause much trouble.

Well any USB device that supports remote wakeup capability. For example if you have Logitech dongle connected it could cause it too.

But the most important part is that you can turn it off if you want to have something connected.

[mb_q]

Does it mean there are AMD laptops supporting S3 rather than "modern standby"/S0ix?

[superm1]

Does it mean there are AMD laptops supporting S3 rather than "modern standby"/S0ix?

Laptops with Rembrandt, Mendocino, and Phoenix APUs don't support S3. If you have an AMD laptop that supports s0i3 and have problems, please reproduce the issue using scripts/amd_s2idle.py. Attach the report to a bug report at Issues · drm / amd · GitLab and tag me and I'll look at it.

[andrei_me]

All hail superm1 for improving AMD laptops